All posts

Transcribe Videos from Claude, Cursor & Windsurf with MCP

April 16, 20265 min read

You're in Claude, Cursor, or Windsurf working on a project. Someone drops a TikTok link in Slack and says "can you pull the transcript from this?" You open a new tab, paste the URL into a transcription tool, wait, copy the result, paste it back. Four context switches for one piece of text.

The ReelScribe MCP server eliminates all of that. Paste the video URL into your AI chat, and the transcript comes back in the same conversation — no tabs, no copy-pasting, no leaving your editor.

What Is MCP?

MCP (Model Context Protocol) is an open standard created by Anthropic that lets AI assistants connect to external tools. Think of it as a plugin system — the AI assistant discovers what tools are available, and uses them when relevant. The protocol is supported by Claude Desktop, Claude Code, Cursor, Windsurf, Cline, and a growing number of other clients.

The ReelScribe MCP server is an npm package that wraps the ReelScribe API and exposes it as a set of MCP tools. Once configured, your AI assistant can transcribe videos, check your credit balance, and search your transcription history — all through natural language.

Setup (2 Minutes)

1. Get an API Key

Sign up for free (25 credits included) and generate an API key from your API settings page. Your key starts with rs_.

2. Add the Server Config

Add the following to your MCP client configuration. The server runs via npx — nothing to install globally. This config works for Claude Desktop, Claude Code, Cursor, and Windsurf:

MCP Configuration — claude_desktop_config.json / .mcp.json
{
  "mcpServers": {
    "reelscribe": {
      "command": "npx",
      "args": ["-y", "reelscribe-mcp-server"],
      "env": {
        "REELSCRIBE_API_KEY": "rs_your_api_key_here"
      }
    }
  }
}

File locations: Claude Desktop uses ~/Library/Application Support/Claude/claude_desktop_config.json, Claude Code uses a project .mcp.json or global ~/.claude.json, and Cursor/Windsurf use their Settings > MCP panel.

3. Start Using It

Restart your AI client. The ReelScribe tools are now available. Try asking:

"Transcribe this video: https://www.tiktok.com/@user/video/123456"

The server submits the URL, polls for the result, and returns the full transcript — all within a few seconds. No need to learn tool names or pass JSON; the AI figures out which tool to call based on your message.

Available Tools

The MCP server exposes six tools. Here's what each one does and when your AI assistant will use it:

  • transcribe_video — Submits a URL and waits for the transcript. Costs 1 credit. Example: "Transcribe this TikTok: [url]"
  • get_transcription — Retrieves a transcription by ID or request ID. Example: "Get the transcript from request abc123"
  • list_transcriptions — Lists your transcriptions, optionally filtered by status. Example: "Show my completed transcriptions"
  • search_transcriptions — Searches by video URL to check if already transcribed. Example: "Did I already transcribe this? [url]"
  • get_credits — Shows your credit balance, tier, and storage usage. Example: "How many credits do I have?"
  • validate_url — Checks if a URL is supported without using credits. Example: "Is this URL supported? [url]"

What Happens Under the Hood

When you ask your AI assistant to transcribe a video, here's the sequence:

  1. The AI detects the intent and calls transcribe_video with the URL.
  2. The MCP server validates the URL client-side (platform detection, format check) and rejects unsupported URLs immediately — no credit charged.
  3. If valid, the server POSTs to the ReelScribe API. If the video was already transcribed on your account, the existing result is returned instantly (no duplicate charge).
  4. For new transcriptions, the server polls every 3 seconds for up to 90 seconds until the transcript is ready.
  5. The full transcript, metadata (duration, hashtags, mentions, engagement stats), and timestamped segments are returned to the AI.

Most transcriptions complete in 10–30 seconds. If a transcription is still processing after 90 seconds, the server returns the request ID so you can check later with get_transcription.

Supported Platforms

  • Instagram — Reels, Posts (/p/), IGTV (/tv/)
  • TikTok — Standard videos, short links (vm.tiktok.com, vt.tiktok.com)
  • YouTube — Videos, Shorts, short links (youtu.be)

All platforms support 99+ languages with automatic language detection.

Real Use Cases

Content Repurposing in Claude Code

You're writing a blog post and want to reference something someone said in a YouTube video. Instead of finding the timestamp manually, ask Claude to transcribe the video and pull the relevant quote. The transcript stays in context — you can ask follow-up questions, request a summary, or extract key points.

Research in Cursor

Building a feature inspired by a tutorial video? Transcribe the video and ask Cursor to extract the code patterns, architecture decisions, or implementation steps. The transcript becomes part of your coding context.

Social Media Analysis

Monitoring competitors on TikTok or Instagram? Transcribe a batch of their recent videos and ask Claude to identify recurring themes, key messages, and content patterns. Use list_transcriptions to review everything you've transcribed so far.

Pricing

The MCP server itself is free and open source (MIT license). You pay for ReelScribe credits — the same credits used on the web app:

  • Free: 25 credits on signup — no credit card needed
  • Creator: $9.99/mo — 1,000 credits
  • Agency: $24.99/mo — 3,000 credits
  • Enterprise: $59.99/mo — 10,000 credits

Each transcription costs 1 credit regardless of video length or platform. Duplicates are free. See full pricing details.

Frequently Asked Questions

Do I need to install anything globally?

No. The server runs via npx -y reelscribe-mcp-server, which downloads and runs it on the fly. No global installs, no version management.

What if a transcription fails?

Your credit is automatically refunded. The server returns the error details so you (or the AI) can decide what to do next — retry, try a different URL, or move on.

Can I use this alongside the n8n node or the web app?

Yes. The MCP server, n8n community node, and web app all share the same account and credit balance. Transcriptions are shared too — a video transcribed via MCP appears in your web dashboard, and vice versa.

Is it open source?

Yes. The server is published on npm as reelscribe-mcp-server under the MIT license. Source is on GitHub.

Try it now — 25 free credits

Sign up, generate an API key, and add the config to your AI assistant. First transcription in under 2 minutes.

Get Your Free API Key →

Also see: MCP Server landing page · n8n community node guide · Best free video transcription tools